You are on page 1of 9

Home Search Collections Journals About Contact us My IOPscience

Simulation of partial discharges in conducting and non-conducting electrical tree structures

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2001 J. Phys. D: Appl. Phys. 34 1235

(http://iopscience.iop.org/0022-3727/34/8/314)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 132.203.227.62
This content was downloaded on 27/01/2015 at 08:38

Please note that terms and conditions apply.


INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 34 (2001) 1235–1242 www.iop.org/Journals/jd PII: S0022-3727(01)21215-7

Simulation of partial discharges in


conducting and non-conducting electrical
tree structures
J V Champion and S J Dodd
London Guildhall University, 31 Jewry Street, London, EC3N 2EY, UK

Received 22 January 2001


Abstract
Electrical treeing is of interest to the electrical generation, transmission and
distribution industries as it is one of the causes of insulation failure in
electrical machines, switchgear and transformer bushings. Previous
experimental investigations of electrical treeing in epoxy resins have found
evidence that the tree structures formed were either electrically conducting
or non-conducting, depending on whether the epoxy resin was in a flexible
state (above its glass transition temperature) or in the glassy state (below its
glass transition temperature). In this paper we extend an existing model, of
partial discharges within an arbitrarily defined non-conducting electrical tree
structure, to the case of electrical conducting trees. With the inclusion of
tree channel conductivity, the partial discharge model could simulate
successfully the experimentally observed partial discharge activity occurring
in trees grown in both the flexible and glassy epoxy resins. This modelling
highlights a fundamental difference in the mechanism of electrical tree
growth in flexible and glassy epoxy resins. The much lower resistivities of
the tree channels grown in the glassy epoxy resins may be due to conducting
decomposition (carbonized) products condensing on the side walls of the
existing channels, whereas, in the case of non-conducting tree channels,
subsequent discharges within the main branches lead to side-wall erosion
and a consequent widening of the tubules. The differing electrical
characteristics of the tree tubules also have consequences for the
development of diagnostic tools for the early detection of pre-breakdown
phenomena.

1. Introduction is a fundamental difference in the electrical treeing behaviour


dependent on whether the epoxy blend sample was in a glassy
Electrical treeing is of interest to the electrical generation, or flexible state. One possible explanation of this was the
transmission and distribution industries as it is one of the causes very different value of the elastic modulus for the epoxy
of insulation failure in electrical machines, switchgear and blends above and below their glass transition temperature.
transformer bushings. The process can be split into two distinct However, other experiments [3] have shown that the electrical
phases [1]. (i) Tree initiation where an initial void is formed characteristics of the resultant tree structures were significantly
at a region of high electrical stress within the insulation. This different. In particular, electrically conducting trees were
is followed by (ii), the growth of branch structures (electrical found to form in epoxy resins in the glassy state (at test
trees) consisting of gas filled tubules. Electrical discharges temperatures below their Tg ) and non-conducting trees formed
(partial discharges) within the tree structure drive the growth in epoxy resins in the flexible state (at test temperatures above
process until the tree structure joins the electrodes. Large their Tg ). By conducting we mean that the time constant,
destructive currents can then flow, leading to the breakdown RC, (where R is the effective resistance of the tree segments
of the insulation. and C is the tree–electrode capacitance) is much less than the
Recent experimental work [2] on tree growth in epoxy period of the alternating voltage applied to the electrodes of
resin blends (each blend having a glass transition temperature the sample. All points on a conducting tree therefore have
(Tg ) within the range −8 ◦ C to 50 ◦ C) has shown that there the same electric potential as the applied voltage connected to

0022-3727/01/081235+08$30.00 © 2001 IOP Publishing Ltd Printed in the UK 1235


J V Champion and S J Dodd

Figure 1. Microscopic observations by reflected light of part of a conducting tree structure grown in the glassy epoxy resin CY1301. The
images show the same part of the tree but with slightly different focus planes.

the pin electrode and hence conduction suppresses electrical


discharges from occurring within the main body of the tree
structure. Conversely, non-conducting electrical trees support
partial discharges within the main body of the tree.
Much work has been undertaken on detecting the presence
of electrical trees in high-voltage insulation systems [4] and
diagnostic techniques have been developed [5] which rely on
measuring the current pulses in the external circuit due to
partial discharge events occurring within the tree structure.
These current pulses arise due to a step change in the induced
charge on the electrodes of the insulator following each partial
discharge event. The different electrical characteristics of the
tree structures have been shown [3] to influence the spatial
distribution of partial discharges within the tree structure
and hence will have an impact on the partial discharge
characteristics detected in the external circuit.
We have recently shown [6] that a partial discharge model
based on simple electrostatics and a concept of local electron
avalanches was successful in simulating partial discharges in Figure 2. Microscopic observations by reflected light of a
non-conducting tree structures. In this paper we adapt the non-conducting tree grown in the flexible epoxy resin CY1311.
model to incorporate tree channel conductivity so that the
partial discharge activity in trees grown in the glassy epoxy
resins may also be simulated. A capacitor bridge (the technique is described in [3, 6])
has been used to measure the induced charge on the electrode
of pin–plane samples during tree growth. A typical bridge
2. Experimental evidence for the existence of output voltage over one cycle of the applied voltage is shown in
conducting and non-conducting electrical trees figure 3(a) for non-conducting trees, as growth in the flexible
Visual and microscopic examination of electrical trees under epoxy resins or polyethylene. The bridge signal consists of
reflected light, in glassy and flexible epoxy resins, have shown a stepped oscillating waveform that lags behind the applied
that conducting trees appear black and non-conducting trees voltage. Each step corresponds to a partial discharge event of
appear white. A reflection micrograph of part of a conducting a magnitude equal to the step change in the induced charge
tree, grown in pure CY1301 (CIBA) epoxy resin (Tg = 55 ◦ C) on the electrodes and flows as a current pulse in the external
at room temperature (20 ◦ C) is shown in figure 1. The main circuit. The phase lag of the capacitor bridge signal shows
channel sections appear black and may be due to carbonized that the tree structure was non-conducting. A typical bridge
degradation products. A typical non-conducting tree grown in output signal is shown in figure 3(b) for the case of conducting
pure CY1311 (CIBA) epoxy resin (Tg = −8 ◦ C) is shown in trees as grown in the glassy epoxy resins CT1200, CY1300 and
figure 2. The tree channels appear white by reflected light and CY1301 (CIBA). The bridge output voltage was found to have
the individual tubules of the tree structure appear to be eroded a similar wave shape as the applied voltage and was in phase
smoothly by continued partial discharge activity. with the applied voltage waveform. No step changes were

1236
Conducting and non-conducting electrical trees

the partial discharge rates, i.e. number per half-cycle, did


not increase significantly during the growth of the tree [3].
For conducting trees the partial discharge amplitudes were
much lower and a Faraday cage was necessary to reduce
the noise level of our detection system to below 10 fC. By
using a threshold detector with a variable voltage threshold
the amplitudes of the partial discharge current pulses could be
characterized. It was found that the distribution of amplitudes
of the partial discharges remained in the range 10 fC–10 pC,
but their rate (number per half-cycle) increased during the
growth of the tree. This was due to the localized nature of the
discharges within the tree structure and the increasing number
of discharge sites at newly formed tree channels as the tree
grew in size.
The evidence for conducting and non-conducting trees
described above was obtained mainly from our experiments
on epoxy resin samples. Parallel evidence also exists in the
literature for polyethylene [7, 8]. Tree growth in polyethylene
has been investigated extensively, as it is the base material in
cable insulation. In this class of material, optical and electrical
measurements of partial discharge activity have shown that the
spatial distribution of emitted light can change suddenly during
the development of the tree [8]. For example, during the growth
of a ‘bush/branch’-type tree, light from partial discharge
activity was initially observed from the whole tree structure at
Figure 3. Capacitor bridge measurements over one cycle of the the pin tip. With time, the light emission region grew, reflecting
applied voltage for (a) a non-conducting tree grown in polyethylene the growth of a non-conducting bush tree. After a critical time,
and (b) a conducting tree grown in the glassy epoxy resin CY1300.
The broken curve is the applied voltage waveform. the light emission from the bush structure disappeared and
localized light emission occurred from growing tree tips. No
observed in the waveform, indicating the absence of partial further growth occurred in the bush tree structure; however,
discharges having magnitudes greater than 3 pC, the resolution branch type tree growth started at the periphery of the bush
of the digital oscilloscope used to capture the waveform. An in- tree structure. Coupled with this changeover in the light
phase signal shows that the conducting tree causes the overall emission and tree growth behaviour was a large reduction in
capacitance of the pin–plane sample to increase and hence the the amplitudes of the partial discharge pulses detected in the
tree acts as an electrical extension of the pin electrode. external circuit. The authors [8] concluded that the changeover
The suppression of the partial discharge activity within occurred because the bush structure and the subsequent branch
the main channels of conducting trees can be observed using structure had become conducting.
a sensitive charge-coupled device (CCD) camera to record the
faint light emission from partial discharges. CCD images of the 3. Partial discharge model
light emission from partial discharges within non-conducting
trees show that partial discharges occurred within the main In this section a brief summary of the existing partial discharge
channels of the tree, as shown in figures 4(a) and (b). In the case model will be described. Reference should be made to
of electrically conducting trees, the partial discharges occurred our previous paper [6], which describes the model in detail.
at isolated points at the ends of the growing tree structure, as The modifications to the model to incorporate tree channel
shown in figures 4(c) and (d). The different spatial distribution conductivity will then follow.
of partial discharge activity, within tree structures grown in the
flexible and glassy epoxy resins, will influence the nature of 3.1. Existing partial discharge model
the current pulses flowing in the external circuit. This will
therefore have an impact on the detection of electrical treeing This model was based on simple electrostatics and a concept
in insulation systems. of local electron avalanches to model partial discharges within
Partial discharge measurements, i.e. measurements of the an arbitrarily defined non-conducting electrical tree structure.
current flow in the external circuit, have shown that in the case The electrostatic boundary conditions were applied at all times,
of non-conducting trees [3] the amplitudes of the individual enabling the calculation of the experimentally measurable
partial discharge pulses range from 1 pC to 1 nC and generally quantities of charge flow in the external circuit (partial
increase in magnitude as the size of the tree increases. Typical discharge pulses) and induced electrode charge (capacitor
partial discharge activity over one cycle of the applied voltage bridge measurements) to be modelled. By assuming that
for a 0.6 mm long electrical tree is shown in figure 5. The the light emitted from a local region of the tree structure
partial discharge current pulses predominantly occur in the was proportional to the charge formed locally by the partial
first positive and first negative quadrant of the applied voltage. discharges, the spatial distribution of emitted light could also
Except for periodic bursts in the partial discharge activity, be modelled.

1237
J V Champion and S J Dodd

Figure 4. Sensitive CCD images of (a) a back-illuminated image of a non-conducting tree grown in the flexible epoxy resin CY1311, (b) the
light emission from the tree with the back illumination switched off, (c) a back illuminated image of a conducting tree grown in the glassy
epoxy resin CY1301 and (d) the corresponding light emission image.

Figure 5. Partial discharge pulses measured in the external circuit


over one cycle of the applied voltage for a non-conducting tree
grown in the flexible epoxy resin CY1311.

3.1.1. Electrostatics. A pin–plane electrode arrangement


and the electrical tree were defined on a 45 × 45 square grid
of grid spacing h equal to 50 µm and with a pin tip–plane
spacing of 2 mm. The arbitrary tree structure that was used in Figure 6. An arbitrarily defined tree structure defined on a flat
the model is shown in figure 6. The tree channels were assumed square grid of grid spacing h = 50 µm. The whole 45 × 45 grid is
to be non-conducting and have the same relative permittivity, not shown for clarity.
εr , as the host resin. The tree structure was defined in two
dimensions for convenience. The applied voltage at the pin tip comprised of three contributions: (i) the applied voltage, V (t);
was assumed to be a 50 Hz sinusoidal voltage of peak value
(ii) the potential of the induced image charge, Vim (t); and
V0 . At each time step,
t (3600 per cycle were chosen to give
(iii) the potential at the pin tip position due to charge within the
a phase resolution of 0.1◦ ), the potential within the tree and
tree structure, VQ (t). Thus, to satisfy the boundary conditions
at the pin tip were calculated. All charges in the model were
represented as spheres of charge of uniform charge density and the contributions from Vim (t) and VQ (t) must be equal and
radius h/2. The pin electrode was represented by a sphere of opposite at all times. The equations required to calculate these
charge at the position of the pin tip having a total charge Qapp . contributions are given in [6]. The induced charge on the pin
The plane electrode was represented as the zero equipotential electrode, Qim , due to charge within the tree structure was
surface when an image of the pin tip was placed in its mirror also assumed to occupy a sphere at the position of the pin
image position as shown in figure 7. The boundary conditions tip. Induced charges on the plane electrode were modelled by
were satisfied when the voltage at the pin electrode, Vpin (t), including an image of both the tree structure and the induced
was always equal to the applied voltage, V (t), with the pin charge at the pin electrode in their mirror image positions on
image at potential −V (t). The potential at the pin tip, Vpin (t), the opposite side of the plane electrode.

1238
Conducting and non-conducting electrical trees

Figure 7. Schematic diagram of a partial discharge model electrode and tree structure configuration and the position of the image charges.

potential differences between Verr + Voff and Voff and (ii) all
those tree segments that have not discharges have potential
differences less than Von . The model parameter Verr was a
small error voltage that had a value much less than Von –Voff .
We can make Verr as small as we wish at the expense of
computation time.

Figure 8. The charge formed during a local electron avalanche 3.2. Incorporating tree channel conductivity
called a discharge dipole. The dipole consists of two spheres of
charge of radius equal to half the grid spacing and containing equal In the modified model we assumed that conduction of charge
and opposite charges, ±Q. The charge was assumed to occupy the carriers was ohmic, i.e. that the current flowing along a
spheres with uniform charge density. tree segment was proportional to the local electric field
along the tree segment. All tree segments were assigned a
3.1.2. Partial discharges. A partial discharge event occurring resistance, Rseg , and the current flowing along a tree segment,
at time t was assumed to be made up of one or more iseg = Vseg /Rseg , was equal to
Qseg /
t. However, the
local electron avalanches occurring essentially simultaneously voltage across the tree segment, Vseg , will change as soon as
within the tree structure. The effect of a local electron any charge flow occurs across the tree segment and hence it
avalanche occurring over a distance h (the grid spacing) was to was necessary to apply a relaxation method as described below.
produce a dipole of charge, called a discharge dipole, as shown Each time step in the model,
t, was subdivided into
in figure 8. The discharge dipole consists of charge +Q and a number of sub-time steps,
ts . The number of sub-time
−Q occupying spheres of uniform space charge density and steps, n, was chosen to ensure convergence of the conduction
separation h. The potential difference between the centres of calculation. At each sub-time step, a charge,
Qseg , flows
the two spheres comprising the dipole is
V . The corrections along each tree segment and was calculated from
for induced electrode charge that are made to
V , when the

Qseg = (Vseg /Rseg )
ts .
discharge dipole is near to the pin or plane electrodes, are given
in [6]. The charge
Qseg was assumed to form a dipole of charge
Partial discharges were modelled at each time step, t, by along the tree segment having the same geometry as the
adding one or more discharge dipoles to the tree structure, each discharge dipole shown in figure 8. When dipoles of charge
reducing the potential difference along these segments of the have been added to all tree segments, the electric potential
tree structure that exceed a critical value, Von , to a lower value, within the tree was recalculated and the whole procedure
Voff . If no tree segment had a potential difference greater repeated a further n − 1 sub-time steps to make up one full
than Von , then no partial discharges were possible at this time. time step,
t. It was found in practice that 20 sub-time steps
The model parameter Von was regarded as the critical potential were required for convergence.
difference for electron detachment from gas molecules within
the tree channels and which form the initial electrons for the
3.3. Computational procedure
avalanche process. The model parameter Voff was regarded
as the critical potential difference where the gas ionization The position of the pin and the pin image and the tree structure
coefficient was equal to the electron attachment coefficient and its image were defined on a square grid of spacing h. For
and hence electron avalanches were extinguished below this each tree segment, defined as the line joining two adjacent tree
potential difference [4]. As the addition of each discharge points, values of the parameters Von and Voff were assigned.
dipole to the tree structure changes the induced charge on the At each time step, the electric potential within the tree
electrodes and the electric potential distribution within the tree structure was calculated. A test was then made to see if
structure, it was necessary to employ a relaxation technique for a partial discharge could occur. The maximum segment
adding discharge dipoles to the tree structure at a given time potential difference was calculated. If this potential was
step. The details of the relaxation method used are given in greater than the parameter Von for that segment, then one
[6]. The aim was to keep adding discharge dipoles to the tree or more discharge dipoles were added to the tree structure
structure and recalculating both the image charge and electric using the relaxation procedure described in section 3.1.2. The
potential until: (i) all tree segments that have discharged have addition of all the discharge dipoles was assumed to occur

1239
J V Champion and S J Dodd

Figure 9. Partial discharge model calculations using Rseg = 1 × 1012 : (a) the partial discharge pulses and the capacitor bridge (image
charge) signal over one cycle of the applied voltage (shown as a broken curve), (b) tree structure and (c) distribution of emitted light.

over a time interval much less than the time step used in the current pulses measured in the external circuit, only the first
model. The end of a partial discharge event at time t occurred contribution (i.e. that due to partial discharges) to
Qim (tn )
when no more discharge dipoles could be added to the tree was recorded as a function of time. The spatial distribution
structure. Following the partial discharge calculation, the of light emitted from the tree structure was calculated by
electric potential within the tree structure was then recalculated assuming that the light generated at each tree point on the
and the conduction calculation described in section 3.2 was grid was proportional to the charge generated by the electron
performed. The conduction calculation was always performed avalanches that occurred at each tree point and averaged over
at each time step whether or not a partial discharge event one cycle.
occurred. As tree channel conduction reduces the potential
differences along all tree segments in the tree structure, it was 4. Model calculations
not necessary to retest for partial discharges at the present time
step. Once the conduction calculation was complete, the model The model parameters and their values that were used for the
moves onto the next time step. simulations are shown in table 1. The arbitrarily defined tree
An initial condition of the model was that no space charge structure that was used is shown in figure 6. The applied
existed within the tree structure at time t = 0. Hence, the voltage, V0 , and the model parameters Von , Voff , Verr and εr
calculation required one cycle to overcome the initial transient (the relative permittivity of the CY1311 epoxy resin [9]) were
with data being obtained during the second cycle. The induced kept at fixed values. Model calculations were performed over a
charge on the electrodes Qim (t), due to charge within the tree range of tree segment resistances, Rseg . In figure 9 the induced
structure, was calculated at each time step so that the partial charge on the electrodes, the current pulses in the external
discharge model could be compared directly with experimental circuit and the light emission over one cycle of the applied
capacitor bridge measurements. The model also calculated the voltage are shown for a tree segment resistance of 1 × 1012 .
change in induced charge on the electrodes at each time step, The model calculations, using a tree segment resistance of

Qim (tn ), equal to Qim (tn )−Qim (tn−1 ).


Qim (tn ) comprised 1 × 108  for all segments, produced no partial discharges and
two contributions, (i) an essentially instantaneous change in hence no emitted light as the whole of the tree was sufficiently
induced charge following a partial discharge and (ii) a change conducting to suppress discharges. For the model to produce
in the image charge due to conduction within the tree structure partial discharges within segments at the tree tips, as observed
over the time step
t. In order that a direct comparison experimentally, it was necessary to increase the resistance of
could be made with experimental partial discharge data of the segments at the tree tips to 1 × 1012 . The results for

1240
Conducting and non-conducting electrical trees

Figure 10. Partial discharge model calculations using Rseg = 1 × 108  for the established channels and Rseg = 1 × 1012  for the newly
formed channels at the ends of the tree structure: (a) the partial discharge pulses and the capacitor bridge (image charge) signal over one
cycle of the applied voltage (shown as a broken curve), (b) tree structure and (c) distribution of emitted light.

Table 1. Values for the model parameters. All model parameters The simulated partial discharges occur within the main body
remained at fixed values except Rscg which varied over the range of the tree structure, leading to step changes in the induced
given.
charge on the electrodes and partial discharge pulses of up to
Parameter Comment (unit) Value
60 pC flowing in the external circuit. For the case of conducting
trees it was found that the experimental capacitor bridge (image
V0 Applied voltage (V) 14 142
f Frequency (Hz) 50
charge) and light emission data were similar to that calculated
Von Segment discharge potential (V) 1500 using the partial discharge model for tree segment resistance of
Voff Segment extinction potential (V) 1000 1×108  for the established channels and 1×1012  for the tree
Veff Small error voltage (V) 10
εr Relative permittivity of resin 4.8
segments at the tree tips. The localized partial discharges at the
Rseg Tree segment resistance () 1 × 108 –1 × 1012 end of the tree structure give rise to only small (less than 40 fc)
changes in the induced electrode charge and hence give rise to
much lower magnitude partial discharge current pulses flowing
this non-uniform distribution of tree segment resistances (tree in the external circuit when compared with conducting trees.
with segment resistances of 1 × 108  for the tree segments
The effect of decreasing the tree segment resistance from
comprising the main body of the tree and 1 × 1012  for the
segments at the tree tips) are shown in figure 10. 1 × 1012  to 1 × 108  was to suppress the partial discharges
within the tree structure. Discharges at the tree tips could only
be simulated provided the resistances of these segments were
5. Comparison with experimental data sufficiently high to support partial discharges. The amplitudes
of the modelled partial discharge current pulses drop from
The modelled data can be directly compared with the real
approximately 60 pC to 20 fC as the resistance of the tree
experimental data described in section 2. Close agreement was
found between the modelled data and experimental data. The segments in the main body of the tree were reduced. This was
experimental partial discharge data, capacitor bridge (image coupled with an increase in the partial discharge rate (number
charge) data and the spatial distribution of light emission, per half-cycle) and a change in the bridge output signal, from
obtained from non-conducting trees, were similar to the a stepped oscillating waveform lagging the applied voltage to
modelled data for a tree segment resistance of 1 × 1012 . an in-phase sinusoidal voltage.

1241
J V Champion and S J Dodd

Table 2. Approximate bulk resistivities of the tree segments and rather than gas filled tubules. However, for non-conducting
host resin. tree channels, subsequent discharges within the main branches
Tree channel resistivity Host resin resistivity of the tree lead to side wall erosion and a consequent widening
Resin ( m) ( m) of the tubules. Carbonized deposits do not build up and the
Flexible >2 × 106 1 × 109 –1 × 1011 tree appears white by reflected light. The reason for the change
Glassy <2 × 102 1 × 1012 –1 × 1014 over in the electrical treeing behaviour when the temperature
of the resin is above or below its glass transition temperature is
not known. However, it may be partly due to the very different
6. Discussion mechanical properties of the epoxy resin as characterized by
its mechanical modulus on either side of its glass transition
Charge transport in trees grown in flexible epoxy resin and temperature. However, there is evidence in the literature that
polyethylene was primarily due to partial discharges occurring the changeover between non-conducting and conducting trees
from the pin into the tree structure. The transport of charge also occurs in polyethylene, in this case after a critical time
in trees grown in the glassy epoxy resins was dominated by during tree growth [7, 8].
conduction. The conduction current ensures that the potential The differing electrical characteristics of the tree tubules
gradients along all tree segments were less than that required have consequences for the development of diagnostic tools for
to form an electron avalanche. To form discharges at the the early detection of pre-breakdown phenomena. Detecting
ends of the tree structure it was necessary to increase the electrical treeing in insulation systems made from glassy
segment resistivity to a much higher value. As these segments epoxy resins will be much more difficult owing to the much
were at the tree tips, they can be considered as newly formed smaller magnitude partial discharge pulses in the external
channels in which carbonized degradation products had not yet circuit.
condensed. Each of these tree segments act as a separate site
for partial discharge activity with the remaining tree structure
acting as an extension of the pin tip at the 50 Hz power
Acknowledgment
frequency. The much smaller amplitude partial discharge
The authors acknowledge the support of the National Grid
pulses for conducting trees was due to the much smaller change
Company plc during the course of this work and their
in induced charge on the electrodes following an isolated
permission to publish this paper.
discharge at the end of the tree structure compared with
multiple discharging segments occurring within the main body
of a non-conducting tree. References
If we assume that a realistic size of a tree channel is
approximately 50 µm long and having cross sectional area [1] Dissado L A and Fothergill J C 1992 Electrical Degradation
and Breakdown in Polymers ed G C Stevens (London:
of 10 µm by 10 µm, the tree channel resistances used in the Peregrinus)
model can be converted to bulk resistivities. In table 2 the bulk [2] Champion J V, Dodd S J, Alison J and Askew L 1998
resistivities of the tree channels from the model are compared Electrical tree growth and space charge behaviour of epoxy
with the values for the host resins [9]. For the case of non- resin blends IEEE 6th Int. Conf. on Conduction and
conducting trees grown in the flexible epoxy resin, the tree Breakdown in Solid Dielectrics (22–25 June, 1998), IEEE
Cat. No. 98CH36132, 321-324.
channel resistivities are at most five orders of magnitude less [3] Champion J V and Dodd S J 1998 Charge dynamics during
than the host resin. In the glassy epoxy resins, the tree channels electrical treeing Space Charge in Solid Dielectrics
have resistivities at least 10 orders of magnitude less than the ed J C Fothergill and L A Dissado (UK: Dielectric Society,
host resin. In all cases the bulk resistivity of the tree structure Department of Engineering, University of Leicester)
was much less than the bulk resistivity of the host resin. [4] Van Brunt R J 1991 Stochastic properties of partial discharge
phenomena IEEE Trans. Electr. Insul. 26 902–48
[5] Fruth B and Niemayer L 1992 The importance of statistical
7. Concluding remarks characteristics of partial discharge data IEEE Trans. Electr.
Insul. 27 60–9
With the inclusion of tree channel conductivity, the partial [6] Champion J V and Dodd S J 1998 An approach to the
modelling of partial discharges in electrical trees J. Phys. D:
discharge model can simulate partial discharge activity Appl. Phys. 31 2305–14
occurring in trees grown in both the flexible and glassy epoxy [7] Densley R J 1979 An investigation into the growth of electrical
resins. This modelling highlights a fundamental difference in trees in XLPE cable insulation IEEE Trans. Electr. Insul. 14
the electrical properties of the trees formed in these two epoxy 148–58
resins. The much lower resistivities of the tree channels grown [8] Laurent C and Mayoux C 1980 Analysis of the propagation of
electrical treeing using optical and electronic methods IEEE
in the glassy epoxy may be due to conducting decomposition Trans. Electr. Insul. 15 33–42
products condensing on the side walls of the existing channels, [9] Ciba Geigy 1982 Data sheet CY1311
or that the tree channels are composed of carbonized tracks Ciba Geigy 1982 Data sheet CY1301

1242

You might also like